[gmx-users] Simulated annealing 10000K

[Marc Ceruso]

Hi-
 Your results are not surprizing. But it seems to me that you would be
happier if you had one single structure for you peptide. Even if it were
that the peptide does have a unique stable conformation the search for it
would be to find a protocol/forcefield that would accurately determine the
correct fold of your peptide: protein folding problem. However, small
peptides do not have a unique
structure in solution. Numerous NMR studies on many peptides have shown
that the peptides (in general) are in  equilibirium between different
forms. Take a look at the enkephalin (Scheraga) studies, or some of Ernst
(NMR) studies.
If you have experimental data on your peptide, You could run classical
MD simulations (see Daura and van Gunsteren) for a very long time
(tens/hundreds of ns)and  evaluate
your computational results against the NMR data. That would be the measure
of convergence.
I hope this helps
Marco



  On Thu, 31 Jul 2003, Osmany
Guirola Cruz wrote:

> i have 10 simulations , i do g_cluster to each one and then i do g_confrm
> whith the most populated clusters
> and , these is the table:
>
>          1                  2                 3                4
>          5                 6               7                    8
> 9        10
>  1      *
>  2  0.255333      *
>  3  0.384980  0.268161      *
>  4  0.363748  0.356637  0.312702      *
>  5  0.364454  0.291378  0.168395  0.360910      *
>  6  0.304329  0.362637  0.352839  0.224768  0.350295      *
>  7  0.395214  0.427625  0.373925  0.340693  0.329041  0.264797      *
>  8  0.395720  0.347483  0.315446  0.462918  0.295911  0.384215  0.404703
>  9     ?          ?         ?         ?         ?         ?         ?
> 10  0.269739  0.224798  0.375789  0.433382  0.349531  0.403248  0.452217
> 0.382978
>
>
>
>
> Kay Gottschalk wrote:
>       I'd do much more than ten structures. Do as many as you can
>       in a reasonable time and afterwards cluster the results! You
>       cannot expect your system to converge so easily...
>       On Thursday, July 31, 2003, at 02:00 AM, Osmany Guirola Cruz
>       wrote:
>
>             that was the problem in my simulation at 1000K my
>             ten final structures are quite diferent after the
>             SA.
>
>             Xavier Periole wrote:
>
>
>
>
>
>
>
>             Yop, sorry I meant 1000 K. 10000 K is way too
>             much.
>
>             Marco is right. With 11 residues it is unlikely
>             that your peptide has a unique
>             conformation in solvent. It probably explore
>             different conformations with diffenrent
>             probabilities.
>
>             How did you generate your 10 structures with
>             Modeler. I thought it needed a
>             template !! And what isthe point of doing an SA
>             on a specific conformation ?
>             You loose it at 1000K anyway !!
>             I did some thing similar where I generated 100
>             conf from SA with an implicit
>             solvent (faster) and after that I made clusters
>             of them.
>
>             XAvier
>
>
>
>
>       Dr. Kay-E. Gottschalk
>       Department of Biological Chemistry
>       Weizmann Institute of Science
>       Tel: ++972-8-9343639
>       Herzl St. 1
>       Rehovot 76100
>       Israel
>
>
> _______________________________________________ gmx-users mailing list
> gmx-users at gromacs.org http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the www
> interface or send it to gmx-users-request at gromacs.org.
>